The Dreaded Butterfly Keyboard: trying to postpone the inevitable

In a nutshell: If you have an Apple MacBook Pro or other machine with the dreaded ‘butterfly’ keyboard and it starts to exhibit problems, you options are:

• If you experience the problem of a certain key sometimes not reacting, try detaching the keycap (important: do not try this without first reading the “How to” section below for instructions). After looking—with a magnifying glass—for obvious dust, hairs, or crumbs and carefully removing any with precise tweezers, blast the mechanism with air from multiple directions, and reassemble the key.
• If you experience the problem of a key producing multiple keystrokes per press, first do the above, and after reassembling the key, hit it many times as hard as you're comfortable with. Just type a piece of text like a madman, and there's a good chance that the key will behave afterwards.
• If the above doesn't help, or problems occur too often, or you are not comfortable with taking the keys apart, and you have bought the machine less than 4 years ago, your better option is to go for the Keyboard Service Program.

The full story

A few years ago I was kind of forced to buy a new MacBook Pro, because my previous one, a 2011 model, had the dreaded GPU failure—another design flaw by Apple albeit lesser known than the keyboard debacle. Given the high price and the rumours about the bad keyboard, I seriously considered buying a Thinkpad or other PC laptop, but it looks like I am still too much locked into the Apple ecosystem to take that step. I hoped that the newer design with the extra silicone ‘flaps’ built into the keyboard, would make it reliable enough that it wouldn't develop problems in the situation where the MacBook Pro would be sitting 95% of its lifetime. But that was too much wishful thinking.

Even despite the fact that this laptop sits on a Griffin Elevator in a normal room with regular levels of dust, some keys did develop problems within less than one year. The down arrow key was first, it would sometimes miss a press. Then the same started happening with the Return key. I found a very useful page on iFixit Answers that showed how both thes keys were assembled, and this allowed me to detach them without destroying them. I then blasted the insides of the mechanism with air like crazy, and that has effectively remedied this problem.

A while later however, another problem started showing up: some keys, especially the ‘E,’ started producing double and sometimes even triple presses. The extra press would often come late enough that typing ‘bed’ could result in ‘bede.’ I again took off the ‘E’ key and inspected the internals, but all I found was a tiny hair that was only visible with a 10× magnifying glass. I gave the key the same treatment as the others, but it kept on exhibiting the multiple keystrokes problem.

It seems to me that this key repeating problem is not caused by dust particles. It rather looks like the contacts become unreliable, maybe due to an oxide layer forming on them? The keyboard can become pretty hot when the machine is under heavy load during an extended period, like when running a modern game. Maybe this degrades the contact surfaces? This theory is further corroborated by the fact that I eventually did manage to make this problem go away, by simply punching the key much harder than usual. Percussive maintenance does work, even on delicate modern machines…

The good news is, after a class-action lawsuit, Apple has started a repair program that allows to bring in your machine if it obviously exhibits these problems with the keyboard. They will then replace the keyboard, which usually means replacing the whole top assembly. As explained in this video by Louis Rossmann, theoretically it is possible to replace only the keyboard, but only with very specialised tools because the thing is actually riveted into the case and in many machines you need to detach the glued-in batteries without rupturing them and setting your workshop on fire.

And by the way, I fully agree with Rossman that I expected a machine this expensive to at least be as reliable as a much cheaper one with similar specs, and also much more serviceable. Where are the days when Macs were modular machines that could be entirely taken apart without even needing a screwdriver?

Even though the repair program is free, the downside of course is that you will have to do without your computer for a few weeks. So if you are only having mild problems with the keyboard, it may be a better option to first try to remedy it yourself, but only if you feel confident enough that you won't make things worse.

How to detach the keycaps without destroying them

If you want to have an attempt at detaching the keycaps yourself, BE VERY CAREFUL. Do not simply yank on the caps from any direction, or you will break little hooks and the cap will be forever loose. The trick is to slide a thin piece of plastic in between the cap and the butterfly mechanism (see SHOVE in the pictures below). Do not go too deep, try to keep the plastic as horizontal as possible to ensure it does not go under the butterfly hinges themselves. Then shift the plastic towards the clips (called slide in the pictures) to detach them. For most of the regular keys, the cap is attached with two clips at its top end, and two hooks at the bottom. For these keycaps, start the SHOVE in the middle of the left and/or right edge, as indicated with the arrows in the photo below, then slide upwards. Once one of the clips is detached, you can usually keep sliding across the upper zone of the cap to detach the other clip as well.

A piece of cut-up blister packaging works well for this purpose, but a thin guitar pick (plectrum) should also work well. The longer keys like backspace will typically have 3 clips instead of 2.

For other keys, the procedure differs only in the place where you have to shove the pick and in which direction you should then slide it. For the ‘up’ arrow key, the clips are at the right. For the ‘down’ key, the clips are at the left. For the tall European style of Return key, the clips are at the left, and there are 3 of them. I have borrowed the photos from the iFixit page to illustrate this. Note how the whole mechanism of the return key has been detached in this photo, this is not how it is supposed to be.

To reassemble, slide the hooks back in their place, then re-engage the clips by pressing on the zone where they reside.

The space bar is notoriously difficult. If you feel you need to take it apart, consider going for the Apple repair program instead. If you break things, you won't be able to benefit from the free repair.

And of course, the usefulness of the repair program is very debatable. The replaced keyboard will suffer from the same flaws. I wish there would be a replacement program where I could exchange my inherently flawed MacBook Pro for a considerable rebate on a late 2019 MacBook Pro with its reliable scissor keyboard.

A final piece of advice

If it isn't already obvious: never do anything in the vicinity of your butterfly keyboard equipped machine, that could cause any small particles to fly around. Or, cover the keyboard with a sheet of paper if there is any such risk.

Recently, the backspace key started acting up. Something had obviously got inside the mechanism, but with the naked eye I couldn't see anything suspicious. Even after repeated sessions of removing the key cap and blowing air, it kept feeling strange and sometimes did not respond to presses. I then took a flashlight and magnifying glass, and finally saw that some tiny piece of something had got under the mechanism. It was a pain to remove because Apple's silicone flaps that were supposed to keep junk from getting inside, now made it hard to get the junk out. I eventually managed it with extra fine tweezers, and it proved to be a tiny piece of clipped fingernail any other keyboard wouldn't care about. So, don't clip your fingernails in the vicinity of the keyboard…

Calibrating a cheap digital Chinese hygrometer

This article describes an method that could be used to get reasonably correct read-outs on a digital hygrometer even if it has no provisions for calibration at all. This was mostly an experiment but it has proven to work for the few cases where I tried it. Don't expect miracles from this method, but it may keep the hygrometer readout pretty reliable around the point where it will normally be used.

Introduction

Very cheap combined digital thermometer and hygrometer units can be bought from the usual Chinese marketplaces. The thermometer part of these things is usually reasonably accurate, but the hygrometer part is a game of roulette. Especially if you allowed the sensor to become truly wet by placing the meter in a room with 100% humidity, the risk is high that the calibration will be way off. A proper hygrometer should have a calibration button or button combo that has to be pressed after leaving the unit in an enclosed space with a saturated salt solution, which will stabilise at 75% RH after many hours. These cheap things however rarely have such button.

In the hopes that the device shown in the above photo would have some secret button combo to reset the measurement to 75%, I have tried inserting the battery while holding down buttons in various combinations, or mashing the buttons while the unit was already on, but without any success. I doubt whether devices like this contain any EEPROM at all, the measurement tables are probably hard-coded in ROM, and it is assumed that the humidity sensor will be accurate enough.

Hence the only way to correct the read-out, is to modify the signal coming from the sensor itself. There are two possible approaches:

1. Add resistors to either increase or reduce the electrical impedance of the sensor. This is probably a bad idea however, because it will change the slope of the humidity response curve. The meter will only be correct around a very limited set point, and deviate increasingly when moving away from this set point. It would take some fancy op-amp circuitry to truly shift the impedance curve without altering its slope, and this is way too much effort for a cheap device like this: just buy a better unit for the money that would go into buying parts and spending time on building this circuit!
2. Physically alter the sensor itself. These sensors are simple: it is a substrate sandwiched between two metal plates, or with two intertwined grids of metal ‘fingers’ at one side. The substrate contains some mildly hygroscopic material that will absorb moisture from the air, and the more humid the substrate, the better it conducts electricity and the lower the impedance. Therefore if we can alter the degree to which the substrate wants to absorb water, we can shift the impedance value. I'm not sure whether this won't also alter the slope of the impedance curve, but from my initial tests this seems pretty OK when using the right substances.

I have tried approach 2 with reasonable success. While in theory you could use only a saturated salt solution to check the correctness of your adjustments at 75% RH, it is better to also rely on a true calibrated hygrometer to verify that the cheap meter gives more or less correct readings around the humidity level where you will use it the most.

The method (updated 2020/11/25)

A warning in advance: to get good results, you will need a ton of patience. This is not something you can get right in a few hours unless you are extremely lucky. This takes multiple iterations of a cycle that takes at least 2 days. If time is money, buying a good and reliable meter is much more economical.

I have encountered two styles of sensors and they react somewhat differently to manipulation, but in the end the following strategy seems to work best. We'll start out by applying too much of a substance that is hygroscopic, in other words that likes to absorb water. Then we'll gradually wipe away some of this substance until the read-out is correct. The tricky thing is that wiping the sensor will somehow disturb it in such a way that one needs to wait several days until it stabilises.

Initially I used only hand sanitiser gel as the hygroscopic substance. Such gels will always contain something that attracts water, because otherwise the alcohol in the gel would dehydrate your skin. Typically it will be glycerine (glycerol) but it doesn't really matter what exactly it is. The concentration of this additive is also usually quite low in these gels, which makes it easier to apply small adjustments. The hand gel worked OK for the first type of sensor I tried to tweak. For the second type of sensor however, I had to use a calcium chloride (CaCl₂) solution, mixed with a bit of the sanitiser gel to make it more sticky. I obtained the CaCl₂ from the typical passive humidity absorbers that use a bag or brick of CaCl₂. This absorbs humidity from the air and gradually turns into a brine solution that drips down into a container. These devices are mostly useless as far as dehumidification goes, but they are a good source of CaCl₂. I simply took a drop of the brine solution, mixed it with a bit of the gel, and applied it to the sensor, wiping off the excess such that it wasn't entirely soaked and gave a read-out below 99%.

Next to the hygroscopic substance, you'll also need a classic hygrometer calibration box. This consists of a watertight sealed box, with inside it a small container filled with a saturated salt solution. Take a small cup like a shot glass, fill it with salt, then add water until the salt has become a wet sludge with the water just not spontaneously flowing out of it. This kind of salt sludge has the interesting property that when placing it in a sealed box, it will make the relative humidity inside that box gradually become almost exactly 75%.

So to recap, the things you need are:

• Some hygroscopic substance that sticks to things. What seems to work best is a mix of hand sanitiser and calcium chloride brine.
• A sealed box with inside it a small cup of saturated salt solution,
• a reasonably well-calibrated hygrometer,
• a lot of time.

Steps

The general strategy is to start with too high a read-out and then gradually wipe the sensor to bring it closer and closer to the correct value. The most tricky thing is not to wipe too much, because then you have to start all over again. Wiping the sensor disturbs it only a little. Re-applying the hygroscopic stuff brings it totally out of whack and it takes many iterations to make it stable after that. Therefore you will want to avoid bringing the sensor to a state of too low read-out. Wipe carefully!

1. Start by ensuring the sensor is clean. If there is any residue on it from a previous failed experiment, remove it with water or rubbing alcohol, whatever works best.
2. Cover the sensor uniformly with the hygroscopic stuff. Most likely this will cause the read-out to jump to 99%. Wipe off the excess until it drops below 99%.
3. Now rely on the calibrated hygrometer to continue wiping the sensor until it is near the actual humidity value. Needless to say, take care not to breathe humid air towards the general direction of either meter. This is the most tricky step. In my case I had to wipe until the read-out was 5 to 10% below the actual value. I can't really explain why, and it is possible you may need to aim for a different offset. This is one of the things that makes this process tedious. It seems the required offset became smaller with each cycle, so you should become more careful as the error on the read-out decreases with each cycle of this procedure.
4. Now put the hygrometer in the calibration box with the salt sludge and leave it alone for at least 36 hours. After that, see what value it settled at and remember this value, then take the hygrometer out of the box and let it settle again for at least 12 hours.
5. Now look at the final value and compare to the known humidity. Ideally you will want to do this at a humidity level where you want the hygrometer to be the most accurate.
6. If the read-out is very close to correct, resist the temptation to get it right to the last percent: most likely you will mess it up and will have to start all over. Just stop and be content with what you have. Remember, it is much harder to correct too low a read-out than too high.
7. Otherwise, if the value is still too high and it also was well above 75% in the box, go back to step 3 and wipe, but take care to make increadingly small corrections.
8. If on the other hand the value is too low, it depends. If it is way too low and the value in the box was also well below 75%, you have over-done the wiping and you will have to go back to step 2 and prepare yourself for several more days of going through the same cycle over and over. If it is just a bit too low but the read-out in the sludge box was very close to 75%, then most likely you will never get a perfect reading from this meter across the whole range. You could try to apply a different mix of hand gel and CaCl₂ in the hopes of getting a better overall response curve, but in the end you might just have to be content with slightly too low readings at lower humidities.

The 75% calibration box has a double purpose. First, it offers an initial test to see if the sensor has any chance of being close to correct. Second, it speeds up the stabilising of the sensor after it has been disturbed by the wiping. Again, I cannot really explain why it behaves like this, but in the end it does provide a good result.

As you can see, this method is tedious and involves a lot of trial and error. This is what makes it mostly useless, unless you're in the middle of a virus pandemic and are looking for things to kill your time and gain some usable measurement devices during this process…

Even if you can perfectly nail the calibration to 1% accuracy, of course you still shouldn't trust these things to be reliable. Also never use these simple hygrometers in environments where relative humidity get near to 100%, because in extreme cases, a drop of water may form on the sensor and disturb the hygroscopic layer. Just buy a good hygrometer that can be calibrated if accuracy and reliability is important.

The State of Thingiverse, End of 2019

In June 2016, I bought a 3D printer on a whim, I only had one real concrete idea for something useful to print. But no worries, because there proved to be a website containing a few millions of free 3D models. Granted, the vast majority of models posted on it proved to be useless junk, but some were either very cool or truly useful. That site was Thingiverse. It was, and still is, the de facto standard for 3D printable model sharing. At that time, the site worked pretty well and had a very active and mostly friendly community. Questions would most often be replied to with useful advice, and if there was a problem with the website, there would at least be an acknowledgement even if the problem was not soon fixed.

The Good

Thingiverse used to be one of the main things that kept feeding my interest in 3D printing. Not just because of the new models coming in every few minutes, but also because of the community, and because it was pretty easy to share my own models, many of which started out as improvements upon someone else's. I am now the co-author of the “Flexi Rex with stronger links,” one of the models a present-day buyer of a new 3D printer seems likely to print as one of their very first attempts. I never anticipated this, I only improved upon an existing model because it broke way too soon when my grandson was playing with it. This is what I like about this community, one can easily take an existing model and improve it, and share this improvement for everyone to enjoy.

The Bad

Today however, things have changed for the worse. The turning point was somewhere in 2017, when one of the main moderators of the website, called ‘glitchpudding’, suddenly left. After this, it became much harder to get any response from whomever was responsible for maintaining the site. At the increasingly rare times that there was some kind of announcement, each time it was from a different person I had never heard of before, as if the previous one had been fired. This would not have been that bad if the website would have maintained its same quality level, but it did not. All kinds of annoying issues started popping up, like the website becoming very slow at times or throwing 500, 501, 502, … HTTP errors at random moments.

Complaining about this seemed to be of no use, because only rarely would there be a response from what might either be a Thingiverse employee or maybe just some random joker — there was no way to verify that whoever was replying on the discussion forums, was an actual Thingiverse / Makerbot employee, it was often a different username, more often without than with the “Thingiverse” badge.

Then it got worse: apparently the nearly invisible Oompa-loompas now running the site, were trying to make certain changes for… reasons. One can only guess when something has changed, because there is barely any communication when this happens. The most obvious sign of a change is an increase in the number of issues. Suddenly all photo previews were broken. Then suddenly they worked again, but any photo that was not in a 4:3 aspect ratio, would be shown distorted. This was not the case before: photos used to be letterboxed and/or cropped in reasonable ways. What was the point of this change? It only made the user experience worse.

Now at the end of 2019, there are finally some advance warnings or notifications that some issue is being worked on. For instance in December there was a site-wide banner telling that there would be “maintenance”. One would expect that this maintenance was aimed at improving the state of the website, but instead when it had ended, we had an extra HTTP error in the 500 range and random 404 errors as well, even on the main page. A website gives a very bad impression if its main page throws a 404 error.

Overall the website in its current state gives a strong impression of lack of professionalism, sometimes downright amateurism. It seems that whoever is maintaining it, does no more effort in testing their changes than trying something once and then assuming it will always work. I wonder if they have a testing environment at all. Often it feels as if they make changes to the production servers directly. The changes give an impression of lack of skill in developing a modern cloud-based website, rather it looks like trendy hipster frameworks are just being thrown on top of a rickety organically grown base without much vision. I am not saying the (obviously few) developers working on the site are amateurs, but the end result does give that kind of impression. This is likely to be merely because management does not allow the developers to spend enough time to implement things properly, so they are forced to only quickly hack things together.

These are all ‘feelings’ and ‘impressions’ due to the total lack of communication. Only from the recent 504 error pages I could see for instance that they had either thrown ‘openresty’ into the mix, or had already been using it but had now broken something about it. There has been no announcement for this or an explanation why.

I have created my own issue tracker on GitHub just to make all the most obvious problems with the website more visible, in the hopes that this would help the maintainers to decide what to fix next, but it seems to be completely ignored.

The Ugly: my guess at what is going on here

In case you didn't already know, Thingiverse is owned by Makerbot and since 2013, Makerbot is a subsidiary of Stratasys, a company that already marketed 3D printers way before the big 3D printer boom started around the year 2010. Makerbot has changed from a small company selling affordable open source printers, to a big company selling rather expensive walled-garden machines aimed at the education market. I guess most of the original enthusiastic team wanting to change the world (like glitchpudding) have either been fired and replaced by people only interested in milking profits from whatever looks vaguely promising without actually caring about it, or they have transformed into such people themselves.

Thingiverse was part of the original vision of making 3D printing affordable for the home user, a vision that does not really fit with Stratasys, the company selling industrial machines at industrial prices. I guess that at every moment since 2013 when a decision about Thingiverse had to be made, the decision has been biased towards gradually sunsetting the website. Nobody at those companies seems to understand the value of this huge playground that encourages anyone to buy, experiment with, and get familiar with 3D printers. It doesn't matter that no beginner will immediately buy a horribly expensive Makerbot or Stratasys printer: the mere fact that there is a low threshold to gain experience with 3D printers, will increase the chance that these same tinkerers will later on generate profits for those companies. That however is probably way too much thinking-ahead for the average marketeer who has been brainwashed to always take greedy short-term decisions and lodge themselves into a cozy local optimum.

What Really Is Going On

Shortly after writing this article, someone notified me about this blog post: https://xyzdims.com/2019/11/21/misc-formnext-2019-aka-just-too-much-for-one-day/

It contains a part about Makerbot's presence on the Formnext 2019 exhibition. The author had the chance to talk to Jason Chan, who is responsible for Thingiverse. Many of my suspicions are confirmed: only two developers are assigned to the site (and I guess only part-time), and the company greatly underestimates the importance of Thingiverse. There seems to be some commitment to improve it, but again it looks as if the ones holding the bag of money do not share this commitment… Still no concrete indication of what will actually happen to the site.

“But it's free!”

Every time someone posts a complaint about the broken state of the website on the Thingiverse Group forums, there will be replies in the vein of: “but it is free, you have no right to complain.” I disagree. There is no such thing as a free lunch. Everyone who uploads content to the website, somehow invests in it. Some more than others, depending on how much effort they put in crafting the presentation and documentation of their Things. I have invested quite a lot, with about 120 published Things, each with pictures and an extended description. What I am now getting in return, is a pile of issues that make it harder to upload and edit Things, and I have no idea where the site is heading because of the lack of communication from the part of the maintainers. This lack of communication and lack of care to properly test each change, feels very disrespectful, even though only in an indirect manner. It almost makes me feel like an idiot for having put all this effort in my uploads during the years I have been on the site.

There are a few particular users on the Thingiverse groups who will react religiously against any complaint, one of whom has a pretty apt username given his writing style, which makes it seem as if he is drunk (he probably just is). Ignore them, because they are either trolls feeding on the anger, Makerbot employees paid to hold a denial campaign, idiots, or all of the above. None of them upload much of anything, therefore they don't even have any ground to stand on with their claims that the site works perfectly fine.

The content creators are the only reason of existence for Thingiverse. These creators deserve a little more respect than being ignored and being handed increasingly cumbersome tools to upload their content, without any explanation of what is being changed about the website, when, and why. There is no excuse for such poor communication in this era with so many different digital communication methods (and no, Twitter is not really a good communication method). I know it can be done better, because I was pretty content with how the site was maintained and how changes were communicated when I joined it in 2016. I have the feeling that the main breaking point for the website, was when the aforementioned glitchpudding left somewhere in 2017. It seems to have gone downhill ever since. I am not the one who demands infinite progress in everything, but I do expect that when something is good, people do the effort to keep it good.

Solutions, alternatives?

Obviously any sane person witnessing such evolutions on a certain website, would start looking for an alternative. That however proves to be a big problem with Thingiverse: there is no real alternative. I have looked at some other sites, but Thingiverse's biggest trump card is its sheer library of things. No other site comes close to it, therefore it doesn't matter if Makerbot only keeps Thingiverse at a level where it is just usable enough, people will keep coming for the content.

YouMagine looks decent at a first glance: its interface is the most similar to Thingiverse's I have found so far. It is owned by Ultimaker, alhough this is not explicitly mentioned on the site's main page. After trying it out however, it is obvious that YouMagine suffers from the same lack of maintenance, or worse. The ‘blog’ part has not been updated in ages, the featured things remain the same for a very long time, and reporting spam is impossible because the ‘report’ link only points to a dead support e-mail address. The 3D model previews and the ‘assembled view’ feature are mostly broken. Links and bold text do not work in the description text, which is incredibly annoying (I can imagine that they disabled links due to the spam, but bold text: why?!)
The site still runs, but it looks like a ghost ship. Maybe the only good thing about this is that if nobody changes anything about it, they also cannot break anything about it…

There are other sites like MyMiniFactory, against which I have grown an aversion due to the apparent shills promoting it on Thingiverse groups as the best thing since sliced bread. There is also Cults3D but just as with MMF, I find that there is too much emphasis on hiding content behind a paywall.

I have no concrete ideas for a solution. The best thing would be if a new website would be constructed by a community that does care about it, not tied to one particular manufacturer, and with a strategy to keep the community and website alive in the long term. Ideally, somehow the Thingiverse library would then be migrated to this site, but that is optional. I have made backups of all my uploaded models and their descriptions and photos, and I will happily re-upload them to a new website that is worth it. Maybe it would be better this way because to be honest, I estimate 80% of all Thingiverse models to be junk that would better be garbage collected. Starting from a clean slate might be better…

A ‘solar’ USB power bank with a fake solar panel, really?

A while ago I bought a ‘solar USB power bank’ from an eBay seller, with the idea that this thing would charge itself by merely sitting on a windowsill, at least cutting down on the time to recharge it.

When I received it, I immediately got a bad feeling about it. The solar panel looked weird. However, an LED on the power bank did react to the amount of incoming light, so I assumed it was just a strange new manufacturing process for solar panels. It did however not take long before I took the whole thing apart and found out that it was actually an elaborate scam: the panel was fake, just a piece of plastic with a sensor in it.

On my main website I explain in detail how this scam is designed, how you can recognise it if you suspect you also have a fake solar power bank in your hands, and how I eventually converted this scam into a real working solar power bank by adding a real solar panel and the necessary components. Buyer beware!

Read the full article

Een ‘solar’ USB powerbank met namaak zonnecel, echt?

Een tijdje geleden kocht ik een ‘solar USB powerbank’ van een eBay-verkoper, met de idee dat dit ding zichzelf zou opladen door louter op een vensterbank te liggen, en zo op zijn minst de tijd zou verkorten om hem helemaal op te laden.

Toen ik het ding ontvangen had, had ik meteen in de mot dat er iets niet pluis was. Het zonnepaneel zag er vreemd uit. Echter, een LED op de powerbank reageerde netjes op de hoeveelheid binnenkomend licht, dus nam ik aan dat het gewoon een vreemd nieuw fabricageproces voor zonnepanelen was. Het duurde echter niet lang eer ik het ganse ding toch uiteenhaalde, en erachter kwam dat het inderdaad een hoop vernuftig bedrog was: het paneel was namaak, gewoon een stuk plastiek met een lichtsensor erin.

Op mijn website beschrijf ik in detail hoe dit bedrog ontworpen was, hoe u het kan herkennen als u vermoedt dat u ook zo'n namaak powerbank op zonne-energie in handen heeft, en hoe ik uiteindelijk dit stuk bedrog omgetoverd heb tot een echte powerbank op zonne-energie, door een echt zonnepaneel en de nodige componenten toe te voegen. Caveat emptor!

Lees het volledige artikel (Engels)

Dark Sky / Forecast.io: avoid using the temperature, humidity, or pressure values of the first hourly block

Here is yet another post about something highly specific that is totally unrelated to everything else on this blog. Enjoy!

If you rely on Forecast.io to obtain weather prediction data, you may also have implemented fancy things like a graph of how the temperature will evolve over the next few hours. Or maybe just a rising/falling trend indicator. This seems very easy to do: just take the “hourly” data block and take the “temperature” values of each entry, and either plot them or compare them. This will work, but you may notice that as the time gets nearer to the next hour change (eg. 09:57), the first point in your graph starts wobbling around wildly every time you refresh the data. Why is this?

A plot made at 16:55, current reported temperature was 0.1°C.

The first hourly data block represents the current ongoing hour. For instance if you retrieve predictions at 09:57, the first block will represent 09:00. One would expect that the folks at Dark Sky would rely on their historical data to fill in the values for that block, given that it represents a moment in the past. For some reason however they merely take the current observations, in this example from 09:57, and extrapolate them into the past using the data for the next predicted hourly block. For instance if it is currently 15.12 degrees and the temperature for 10:00 is predicted to be 15.79 degrees, then the value for 09:00 will be calculated as (15.12-15.79*57/60)/(1-57/60) = 2.39 degrees.

Extrapolation works fine for data that is known to be relatively stable if the extrapolated point is near the interval of known data. In this case only one point is an actual observation though, the other one is a prediction: also an extrapolation in its own right. This makes it doubly dangerous to extrapolate another value out of it, especially if it lies so far away from the two given points.

The morale of the story is: do not rely on the temperature, humidity, or pressure values from the very first hourly data point, unless the hour has only just started. If you want to make a temperature graph, it makes more sense anyway to start the plot at the current time and not at the start of the current hour. Even if at some point the people at Forecast.io update their system to really put historical data in that block, if you would merely dump the values from all blocks in a graph, visitors would still be seeing a curve that starts almost an hour ago when they load your page 09:57, which is not very intuitive.

The same plot as above, but now starting at the actual observation.

After figuring this out, I now start my plot with the current observation and I calculate the trends by comparing the current observation to a linear interpolation at the current time plus one hour. Interpolating also is not ideal but a lot more defendable here than a wild extrapolation.

Watch Phones, Smartwatches, and the iWatch

Recently some rumours about an Apple “iWatch” have popped up, and many people thought the idea of an internet-enabled communication device on one's wrist is something novel. Not really: so-called watch phones have existed for more than ten years. Of course, the first ones were clunky and barely went beyond prototype stage, but there have been widely available and quite usable watch phones for more than five years. Most of them originate from China, and can be bought in various places for reasonable prices. My main cell phone has been a watch phone since June 2008.

The main problem with pretty much every watch phone that currently exists however, is that it severely lacks in a certain way. And for some reason, whenever a manufacturer produces a new model that improves upon certain features, it gets worse in other features — I dubbed this “the law of conservation of suckiness”.

In a new article on my site, I shortly discuss the history of watch phones, what is good and bad about past and current models, why the concept of smartwatches that need a smartphone makes no sense from an economical point-of-view, and last but not least: a concept of a watch phone which I believe could become popular, and which I would really like to see manufactured in the near future.

One of the biggest problems is that whenever I tell someone that my watch is also a phone, they will immediately ask if it isn't uncomfortable to make calls with it and if I need to hold my arm in silly poses. The answer is: no. Yet, most current models do effectively suffer from this problem. My proposal contains a solution to get rid of this misconception. Watch phones will not become a commercial success as long as people have this Knight Rider-inspired mental image about them.

Uurwerktelefoons, Smartwatches en de iWatch

Recent zijn er geruchten over een Apple “iWatch” opgedoken, en velen dachten dat het idee van een communicatietoestel met internetverbinding dat op de pols gedragen kan worden baanbrekend was. Niet echt: zogenaamde watch phones of uurwerktelefoons bestaan al langer dan tien jaar. Natuurlijk waren de eerste modellen lomp en kwamen nauwelijks voorbij het prototypestadium, maar sinds vijf jaar zijn er wijd verspreide en behoorlijk bruikbare uurwerktelefoons. De meeste hiervan komen uit China, en kunnen op verschillende websites gekocht worden aan democratische prijzen. Sinds mei 2008 was mijn GSM voor dagelijks gebruik een uurwerktelefoon.

Het grootste probleem met elk model uurwerktelefoon dat nu bestaat is echter dat het zwaar tekortschiet op een of meerdere gebieden. En om een of andere reden vinden fabrikanten het nodig om telkens ze iets verbeteren aan een nieuw model, het te verslechteren op andere gebieden. Ik heb dit “de wet van behoud van onbruikbaarheid” gedoopt.

In een nieuw artikel op mijn website geef ik een korte geschiedenis van uurwerktelefoons, beschrijf ik de goede en slechte punten van vroegere en huidige modellen, waarom het idee van smartwatches die een smartphone nodig hebben om te werken op niets slaat, en tenslotte: een concept voor een uurwerktelefoon waarvan ik geloof dat hij wél populair zou kunnen worden, en waarvan ik hoop dat iemand hem kan produceren in de nabije toekomst.

Een van de grootste problemen is dat telkens ik iemand vertel dat mijn uurwerk ook een telefoon is, zij ogenblikkelijk vragen of het niet oncomfortabel is om te bellen en of ik mijn arm niet in een idiote pose moet houden. Het antwoord is: nee. Desondanks lijden de meeste huidige modellen effectief onder dit probleem. Mijn voorstel bevat een oplossing om dit misverstand uit de weg te ruimen. Uurwerktelefoons zullen nooit een commercieel succes worden zolang mensen er dit door Knight Rider geïnspireerde idee over hebben.